GeoStep documentation

GeoStep supports the design and analysis of randomised geographic experiments. Its primary workflow uses two-arm complete or stratified allocation and parallel outcome analysis. Causal interpretation requires a defensible assignment, measurement and execution process.

Start with installation, run the complete Python example, then follow the agency workflow. Read methodology before choosing an estimand or interpreting an interval.

The API reference describes the implemented interfaces. Interpretation, advanced boundaries and troubleshooting cover decisions and failure handling.

Release evidence

The 0.5 release protocol fixes the tests and acceptance criteria. The numerical review, 0.5 simulation results and 0.5 release verification define the private release’s support limits. CRT and staircase inference remain experimental.